Replace values in data frame column with results of aggregate/cut - r

Question

Edit: Modified simulated data so that price means/medians and neighborhoods don't overlap perfectly.
I have a column in a dataframe, we'll call it Price. I'm just simulating data here:
mydata = data.frame(index = rep(1:1000))
mydata$price[1:300] = rnorm(250, mean = 10000, sd = 1000)
mydata$price[301:550] = rnorm(250, mean = 25000, sd = 1000)
mydata$price[551:775] = rnorm(250, mean = 75000, sd = 1000)
mydata$price[776:1000] = rnorm(250, mean = 100000, sd = 1000)
And a set of neighborhoods, we'll call it Hoods:
mydata$hoods = factor(c(rep('hood1',250),rep('hood2',250),rep('hood3',250),rep('hood4',250)))
Then I aggregate the neighborhoods by the median price to create a median bin. I'd like to bin neighborhoods by their median price.
agg <- aggregate(mydata$price, by = list(hoods), FUN = median))
Then I create a cut version of the neighborhood medians (in my actual data there are 24 neighborhoods). So something like:
cut_aggregates <- cut(agg$x, breaks = c(0, 25000, 70000, 110000), labels = c('low','medium','high'))
I then want to replace the value of every 'hood1' in the original data with the aggregated price label, and so on for all the Neighborhoods. SO the first 250 records would be 'low', for example. I know I could make some nested if statement, or brute-force hard-coding. Does anyone know a way I can more efficiently assign all the values in one go, as I may use this for datasets larger than 1000 records. Thank you very much for any help you may provide.
in the final output, the categorized neighborhood ('low', 'medium', 'high') won't necessarily be the same as just doing a cut on price from the original data, because some neighborhood will have a combination of 'low', 'medium', and 'high' using this strategy. I want to first categorize each neighborhood based on its aggregate, and THEN recode the neighborhood.

Answer 1

A very simple way to do this, and probably the fastest, is to use data.table.
library(data.table)
# convert mydata into a data.table
setDT(mydata)
# calculate median price by hood
mydata[, med := median(price), by=hoods]
now you can either:
# replace the original data of `hoods` with the new price labels
mydata[, hoods := cut(med, breaks = c(0, 25000, 70000, 110000), labels = c('low','medium','high'))]
# or create new price labels in a new column
mydata[, new_col := cut(med, breaks = c(0, 25000, 70000, 110000), labels = c('low','medium','high'))]
Finally, if you want just a summary table for each hood:
mydata[, (med = median(price)), by=.(hoods, new_col)]
> hoods my_cut V1
> 1: hood1 low 9916.564
> 2: hood2 low 24696.864
> 3: hood3 high 74749.481
> 4: hood4 high 99852.744

Answer 2

Edit: Approach 1
mydata <- within(mydata, med <- ave(price, hoods, FUN = median) )
mydata$new_label <- cut(mydata$med, breaks = c(0, 25000, 70000, 110000), labels = c('low','medium','high'))
# index price hoods med new_label
# 1 1 10084.756 hood1 10014.38 low
# 2 2 10226.460 hood1 10014.38 low
# 3 3 10432.556 hood1 10014.38 low
# 4 4 10558.065 hood1 10014.38 low
# 5 5 10059.755 hood1 10014.38 low
# 6 6 9885.359 hood1 10014.38 low
Approach2:
Since agg$labs is not unique for each level of hoods in mydata, it will be better to reassign labels individually using a loop by mapping the levels of hoods with the values of agg$labs.
If you had unique labels in agg$labs for each levels of hoods in mydata, then it will be very simple to just reassign labels by doing mydata$hoods <- factor( mydata$hoods, levels = agg$Group.1, labels = agg$labs ). However you have duplicated levels in agg$labs, so you will follow the steps below.
mydata$hoods <- as.character( mydata$hoods ) # convert factor to character
agg$labs <- as.character(agg$labs) # convert factor to character
for( i in seq_len( nrow( agg ) ) ) { # change labels for hoods in mydata
mydata[ mydata$hoods %in% agg$Group.1[ i ], "hoods" ] <- agg$labs[i]
}
mydata$hoods <- factor( mydata$hoods ) # convert hoods back to factor
unique(mydata$hoods) # output
# [1] low medium high
# Levels: high low medium
Data:
set.seed( 200 )
mydata = data.frame(index = rep(1:1000))
mydata$price[1:250] = rnorm(250, mean = 10000, sd = 1000)
mydata$price[251:500] = rnorm(250, mean = 25000, sd = 1000)
mydata$price[501:750] = rnorm(250, mean = 75000, sd = 1000)
mydata$price[751:1000] = rnorm(250, mean = 100000, sd = 1000)
mydata$hoods = factor(c(rep('hood1',250),rep('hood2',250),rep('hood3',250),rep('hood4',250)))
agg <- with(mydata, aggregate( price, by = list(hoods), FUN = median) )
agg$labs <- cut(agg$x, breaks = c(0, 25000, 70000, 110000), labels = c('low','medium','high'))
agg
# Group.1 x labs
# 1 hood1 10014.38 low
# 2 hood2 25021.96 medium
# 3 hood3 74963.40 high
# 4 hood4 100019.88 high
The data in agg will vary if you choose a different seed in set.seed() function.